Discussion

Our exploratory study used the ImageJ software to look at postmortem injuries through quantitative morphometric and gray-level texture analysis. We noticed key differences in parameters such as area, perimeter, solidity, and gray-level texture. This shows that these objective image-based measurements could give extra information for forensic injury assessments and record-keeping.

Digital image analysis is now a key tool in cutting down on observer dependency in forensics. According to Moscalu et al. (2023) [24], computational methods can objectively pull out exact data from biological images, both shape-related and texture-related, that might boost the reliability of results compared to just eyeballing them. This study backs that up, too; the ImageJ tools did a great job at measuring injury details, which showed clear statistical differences between the types of injuries.

In this study, a key finding was that RTA injuries had larger areas and perimeters compared to other types of injuries. This aligns with the findings of Baruah et al. (2025) [25], which showed that quantitative shape features such as area and perimeter help distinguish between different forensic image types. Though they looked at ballistic evidence, the idea of using these measurable traits was pretty much the same. It reinforces how useful morphometric analysis can be in forensic settings.

There is a strong positive link between area and perimeter (r=0.88), matching up with established morphometric rules. Typically, bigger injuries have longer borders, causing those measures to go up together. This connection is backed by earlier quantitative image-analysis studies that found these geometric variables to be pretty interdependent. To check for model instability, they used VIF analysis. Fortunately, all VIF values stayed under 5, meaning the multicollinearity was okay. So, the regression model is good to go.

Maintaining standardization in forensic photography is key to getting accurate measurements. Ferrucci et al. (2016) [26] pointed out that proper calibration scales and set procedures are crucial for forensic photo analysis. For our study, we adhered to these standardized methods and used the ABFO No. 2 forensic scale for all pictures. Doing this probably reduced errors and made our quantitative assessments more reliable. Standardized procedures really matter because changes in camera angle, lighting, or scale placement could seriously affect measurements of both shape and texture.

Among all the variables looked at, solidity showed the biggest differences between injury types and was the strongest predictor in the analysis. Solidity tells us about the regularity and compactness of a shape's edge. If solidity is high, it means the injury edges are smooth and regular. Low solidity, in contrast, suggests rough or broken edges. Patra et al. (2023) [27] pointed out that understanding wound shapes and edges is crucial for forensic work. Our current findings build on this, showing that we can actually measure these features objectively through digital image analysis, instead of just relying on what we see with the naked eye.

Gray-level texture analysis was really helpful in our investigation too. It showed certain key differences among injury groups, and the gray-level SD was statistically significant in logistic regression analysis. These texture parameters show how varied the tissue is inside the region we look at.

Harris et al. (2018) [14] found that texture-based features can effectively highlight structural differences in biological tissues. Wilk et al. (2025) [16] showed that advanced imaging can identify bruises and soft-tissue injuries well. Our results fit with this growing evidence that grayscale texture analysis could very well supplement traditional morphometric measurements for better image interpretation in forensics.

The exploratory logistic regression model found that area, gray-level SD, and solidity are key in figuring out injuries, with solidity showing the strongest connection. Solidity has the highest OR, meaning it associates strongly with injury classification. These results suggest that combining multiple image-based features may improve the differentiation of injury types. Other studies in predictive image analysis came to similar conclusions: combining shape-based and texture-based factors tends to boost classification more than using single parameters. Similarly, Moscalu et al. (2023) [24] highlighted the potential value of multiparametric approaches that integrate multiple image-derived features in predictive image modeling.

The current study also assessed inter-observer agreement and demonstrated high reliability for area, perimeter, and solidity, supporting the reproducibility of the ImageJ-based measurement approach. This consistency is crucial because reproducibility is necessary for any quantitative image-analysis technique to be used in forensics. In summary, the research shows that analyzing shapes and textures gives reliable measurements. These could back up standard ways of examining postmortem injuries. Yet, we should not view these results as solid proof on their own. Instead, they add to the case for using digital image analysis in forensic work. Furthermore, they hint at possible uses in future technology, such as automatic injury classifiers and artificial intelligence (AI) tools that assist forensic experts.

This study has some important limitations to keep in mind. It is exploratory and took place at one location, where categories of injuries were not evenly distributed; there were way more RTA cases than others. With fewer firearm and electrocution injuries, there is less statistical power for a detailed breakdown. Although manual ROI selection had good agreement between observers, it can still vary from person to person. Also, we did not perform external validation, and factors such as postmortem interval, skin pigmentation, and environmental effects could not be fully controlled. So, the results should be considered as preliminary and idea-generating, not as solid guidelines for forensic work. As more than one image depicting injuries may have been obtained from the same autopsy specimen, the conclusions drawn here pertain only to image-level observations.

Future studies should include larger and more balanced datasets with standardized forensic photography protocols. Automated image analysis techniques and external validation studies should be explored to improve reproducibility and generalizability. Additional research using advanced computational approaches may further strengthen the role of quantitative image analysis in forensic medicine.